Deep belief improved bidirectional LSTM for multivariate time series forecasting.

Multivariate time series (MTS) play essential roles in daily life because most real-world time series datasets are multivariate and rich in time-dependent information. Traditional forecasting methods for MTS are time-consuming and filled with complicated limitations. One efficient method being explored within the dynamical systems is the extended short-term memory networks (LSTMs). However, existing MTS models only partially use the hidden spatial relationship as effectively as LSTMs. Shallow LSTMs are inadequate in extracting features from high-dimensional MTS; however, the multilayer bidirectional LSTM (BiLSTM) can learn more MTS features in both directions. This study tries to generate a novel and improved BiLSTM network (DBI-BiLSTM) based on a deep belief network (DBN), bidirectional propagation technique, and a chained structure. The deep structures are constructed by a DBN layer and multiple stacked BiLSTM layers, which increase the feature representation of DBI-BiLSTM and allow for the model to further learn the extended features in two directions. First, the input is processed by DBN to obtain comprehensive features. Then, the known features, divided into clusters based on a global sensitivity analysis method, are used as the inputs of every BiLSTM layer. Meanwhile, the previous outputs of the shallow layer are combined with the clustered features to reconstitute new input signals for the next deep layer. Four experimental real-world time series datasets illustrate our one-step-ahead prediction performance. The simulating results confirm that the DBI-BiLSTM not only outperforms the traditional shallow artificial neural networks (ANNs), deep LSTMs, and some recently improved LSTMs, but also learns more features of the MTS data. As compared with conventional LSTM, the percentage improvement of DBI-BiLSTM on the four MTS datasets is 85.41, 75.47, 61.66 and 30.72%, respectively.

Dual-Specificity Phosphatase 9 Regulates Cellular Proliferation and Predicts Recurrence After Surgery in Hepatocellular Carcinoma.

Hepatocellular carcinoma (CC) is a common and deadly cancer with complex molecular pathogenesis. Little is known about dual-specificity phosphatases (DUSPs) in HCC. We investigated DUSP9 expression in human HCC, associations between DUSP9 and patient outcomes, and effects of altered DUSP9 expression on HCC biology. We studied public data sets as well as 196 patients at our institution who had HCC resections. Quantitative real-time reverse transcription polymerase chain reaction and western blot demonstrated that DUSP9 expression was increased >10-fold in HCC compared to adjacent liver and healthy controls (P = 0.005). Kaplan-Meier and multivariable regression analyses revealed that higher DUSP9 expression was associated with shorter disease-free survival (high DUSP9, 1.6; 95% confidence interval, 0.9-2.3 vs. low DUSP9, 3.4; 95% confidence interval, 1.8-5.0 years; P = 0.04) and increased risk of recurrence (hazard ratio 1.55; 95% confidence interval, 1.01-2.67; P = 0.05) after resection. DUSP9 complementary DNA (cDNA) was cloned using rapid amplification of cDNA ends, revealing two DUSP9 isoforms in human HCC cells. Studies of transcriptional regulation using promoter-luciferase reporter constructs suggested that DUSP9 transcription is regulated by E26 transformation-specific transcription factors. Proliferation of hepatic cells in vitro was enhanced by lentiviral-mediated overexpression of DUSP9. In contrast, DUSP9 knockout HCC cells generated using clustered regularly interspaced short palindromic repeats (CRISPR) demonstrated decreased HCC proliferation and doxorubicin resistance in vitro and impaired xenograft growth in vivo. RNA sequencing, gene set enrichment, and network/pathway analysis revealed that DUSP9 knockout is associated with activation of protein kinase activity and apoptosis. Conclusion: DUSP9 regulates cell proliferation and predicts recurrence after surgery in HCC. DUSP9 may represent a novel prognostic candidate and therapeutic target. Additional studies are warranted to further explore the role and regulation of DUSP9 in HCC.

Differences between anthropometric and bioimpedance measurements of muscle mass in the arm and hand grip and pinch strength in patients with chronic kidney disease.

BACKGROUND & AIMS: Chronic kidney disease (CKD) patients are at increased risk of sarcopenic muscle wasting, with increased mortality risk. Simple screening tests are required to detect sarcopenia to allow for interventional therapies. We wished to compare anthropometric and multifrequency bioimpedance (MFBIA) measurements of arm composition and muscle strength. METHODS: We measured segmental MFBIA, mid arm upper circumference (MUAC) and triceps skin fold thickness (TSF), hand grip strength (HGS) and pinch strength (PS) in CKD patients attending out-patient review. RESULTS: We reviewed 146 patients; 94 male (64.4%), 93 (63.7%) diabetic, mean age 70.5 ± 15 years, weight 77.6 ± 17.1 kg, with a mean HGS of 25.2 ± 10.4, and PS 5.0 ± 1.9 kg. HGS and PS were correlated (r = 0.63, p < 0.001). Male patients had greater HGS and PS (28.3 ± 10.1 vs 19.7 ± 7.0 kg; and 5.3 ± 2.0 vs 4.3 ± 1.1 kg, p < 0.05) with greater arm muscle (3.2 ± 0.7 vs 2.4 ± 0.7 kg, p < 0.05) and less arm fat (1.8 ± 1.3 vs 2.9 ± 1.8 kg, p < 0.05), whereas there was no difference in anthropometric measurements of mid upper arm muscle or fat area. Whereas both HGS and PS correlated positively with MFBIA arm lean mass (r = 0.55, r = 0.37, p < 0.001) and negatively for arm fat mass (r = -0.4, p < 0.001, r = -0.32, p = 0.001) respectively, there were no correlations with anthropometric derived estimates of upper arm muscle or fat. CONCLUSIONS: In CKD patients, segmental MFBIA measurements of the arm, but not those derived from anthropometric measurements demonstrate gender differences and correlate with arm muscle strength, whereas there were no such correlations with anthropometric estimates of upper arm muscle or fat.

Body composition and weakness of hand grip strength and pinch strength in patients with chronic kidney disease from different ethnic backgrounds.

BACKGROUND: Chronic kidney disease (CKD) patients commonly report muscle weakness and fatigue. Losing muscle mass increases mortality. Accordingly, we aimed to determine the main factors associated with loss of muscle mass and muscle weakness. METHODS: Anthropometric measurements were made in CKD patients attending a specialised clinic, along with hand grip strength (HGS), pinch strength (PS) and body composition (muscle mass and fat mass), using segmental bioimpedance assessment. RESULTS: We reviewed the results of 161 CKD patients; 105 male (65.2%), mean (SD) age 70.3 (15) years, body mass index (BMI) 28.8 (6.7) kg m-2 . In multivariable models, both HGS and PS were independently negatively associated with age [standardised ß (St ß) = 0.35; 95% confidence limits (CL) = -0.32 to -0.14; St ß = 0.38; 95% CL = -0.65 to -0.02; P < 0.001, respectively] and positively with appendicular muscle in the arm tested [St ß = 0.34; 95% CL = 2.5-6.3; St ß = 0.24; 95% CL = 0.17-0.98; P < 0.001 and P = 0.006, respectively]. In addition, HGS was associated with male gender (St ß = 0.19; 95% CL = 0.7-7.5; P = 0.019] and negatively with percentage body fat (St ß = 0.22; 95% CL = -0.36 to -0.07; P = 0.003]. There were 47 (29.2%) Asian patients who had lower total skeletal muscle mass/height ratio and appendicular muscle mass/BMI ratio compared to other ethnicities [9.6 (1.8) versus 10.5 (1.6) kg m-2 , P < 0.01; 0.73 (0.23) versus 0.83 (0.33) m2 ; P < 0.01). CONCLUSIONS: In CKD patients, we found that muscle weakness measured by HGS and PS was associated with increasing age and loss of appendicular muscle mass. HGS was also weaker with increasing fat mass and female gender, whereas PS was weaker in patients of Asian ethnicity.

Screening Tests for Sarcopenia in Patients with Chronic Kidney Disease.

BACKGROUND: Patients with chronic kidney disease (CKD) are at increased risk of muscle wasting. Screening tools for sarcopenia, including the Sarc-F questionnaire are now advocated for clinical practice. We wished to compare using the Sarc-F tool with standard measurements of hand grip (HGS) strength and appendicular muscle mass index (APMI). METHODS: We retrospectively reviewed Sarc-F questionnaires completed by patients with CKD, along with contemporaneous measurements of HGS and bioimpedance measured APMI. RESULTS: 146 patients; 94 male (64.4%), mean age 70.5 ± 15 years, body mass index 28.7 ± 6.3 kg/m2 were screened, and 46 screened positive for sarcopenia, with a lower median HGS (19.3 (14.2-24.7) vs 25.6 (19.7-32) kg) and greater body fat (38.3 ± 11.5 vs 30.6 ± 11.5%), both p < 0.001,, with more non-white ethnicity (63 vs 44%), p < 0.05, but there were no other differences. Step-wise adding HGS, and then APMI cut offs, the prevalence of sarcopenia fell from 31.5% to 20.7-24.7% and 2.8-4.8% respectively, with 45.5-62.8% having reduced HGS strength and 11.0-28.1% reduced APMI, depending on which guidelines were applied. Using the most recent European, and ethnicity adjusted cut-off values then there were no statistical differences in the prevalence of sarcopenia with or without the Sarc-F screening tool. CONCLUSIONS: By starting with the Sarc-F screening tool, a number of our patients with CKD would then have been excluded from subsequent investigation for sarcopenia. However, overall screening with the Sarc-F tool did not lead to a significant difference in the prevalence of sarcopenia, when using current and ethnicity adjusted guidelines, compared to combining HGS and APMI alone.

Novel biomarker for hepatocellular carcinoma: high tumoral PLK-4 expression is associated with better prognosis in patients without microvascular invasion.

BACKGROUND: Hepatocellular carcinoma (HCC) recurrence after liver resection (LR) adversely affects prognosis but is difficult to predict. Aberrant expression of Polo-Like Kinase 4 (PLK-4) is implicated in several adult malignancies. We sought to evaluate the prognostic value of PLK-4 expression in HCC after curative-intent LR. METHODS: Patients undergoing LR for HCC between July-2015 and November-2017 at our centre were retrospectively identified. PLK-4 expression was measured in tumour and adjacent non-tumour liver tissue using quantitative RT-PCR. Disease-free survival (DFS) was evaluated by Kaplan-Meier and Cox proportional hazard models. RESULTS: A total of 145 patients were identified. Patients were divided according to PLK-4 expression (high: n = 58, low: n = 87) by generating a receiver operating characteristic curve for recurrence with an area under the curve of 0.72 (95% CI: 0.6-0.8). Recurrence and death rates were similar between groups. In patients without mVI, low PLK-4 expression was associated with worse actuarial DFS (low 1-, 3-, 5-year 83%, 60%, 47% vs. high 91%, 81%, 81%; p = 0.02). In patients without mVI, high PLK-4 expression was an independent predictor of survival (HR 0.3, 95% CI: 0.1-1.0; p = 0.04). CONCLUSION: PLK-4 represents a biomarker for good prognosis in patients with HCC who do not have mVI. This could aid clinical decision making for adjuvant clinical trials.

Comparison of multifrequency bioimpedance measured lean mass to that calculated from anthropomometric measurements in patients with chronic kidney disease.

Patients with chronic kidney disease (CKD) are at increased risk of sarcopenia. Previous studies have proposed equations to estimate muscle mass based on triceps skin-fold thickness and mid-upper arm circumference, with or without adjustment for hand grip strength (HGS). We wished to evaluate their usefulness compared to multifrequency segmental bioimpedance (MFBIA) measured appendicular lean mass (ALM). We audited 160 CKD patients attending outpatient clinics, 65.6% male, median age 73 (62-81.5) years. We calculated muscle mass using six proposed equations based on anthropometric measurements. These equations over estimated muscle mass compared to MFBIA with a mean bias ranging from 3.4 to 35.9 kg. Apart from one equation, there was a systematic bias, with bias increasing with increasing fat mass (ranging from r = 0.17, p = 0.044 to r = 0.65, p < 0.001). For CKD patients we found that most of the previously proposed equations based on anthropometric equations over-estimated muscle mass compared to MFBIA.